credit risk management through securitization: effect on
TRANSCRIPT
Credit Risk Management through Securitization: Effect on Loan
Portfolio Choice
Neil Adrian S. Cabiles Department of Economics University of Bologna
Piazza Scaravilli 2, Bologna Italy Email: [email protected]
Phone: +39 3931424929 Fax: +39 051 209 8143
Abstract
We study how banks take advantage of securitization as a credit risk management
tool. In theory, banks may handle their credit risk exposures by using securitization
to unload risky loans from their balance sheets. Using data on 129 FDIC-member
banks, we have found that while banks indeed use securitization to manage their
credit risk exposures, they do not necessarily do so by totally isolating themselves
from risk. Instead, they exploit the credit risk management property of securitization
to take on greater risk, in pursuit of high returns. As banks carry out this strategy,
the banks’ loan portfolios are complimentarily diversified, giving banks
diversification benefits. These benefits serve as windfall, which, in turn, tempers the
banks’ concerns on the higher risks that they have taken.
JEL Classification: G21, G32, G01
Key words: Securitization, Portfolio Choice, Diversification
2
I. Introduction
Bank engagement in securitization lies on two rationales. One is that securitization
provides the bank with an additional funding source and the other is that it can be a
risk management tool. The former point is explained by the transformation of
otherwise illiquid loans to liquid assets, where these loans are securitized and used
to borrow funds through a Special Purpose Entity (SPE). By this mechanism,
securitization helps banks with their funding matters. In Karaoglu (2005) and
Martin-Oliver & Saurina (2011), for example, it has been observed that banks with
deposit levels that are too low relative to the size of their loan portfolios tend to
securitize their loans, in an effort to address the said issue. Meanwhile, Loustkina
(2011) has found that banks with loan portfolios that are more “securitizable”1hold
less liquid assets2, because the additional funding from securitization makes the
holding of such costly liquidity buffers less necessary.
Alternatively, the risk management role of securitization comes in two forms.
First, which is related to its funding provision, is that securitization can serve as a
hedge against liquidity problems. In Loutskina (2011), this was pointed out in terms
of banks having more securitizable loans in their balance sheets, as being able to
continue and grant more loans even during funding shocks. Similarly, in Cabiles
(2011), we have shown that banks engaging much in securitization can take on more
loan commitments, whose unpredictable takedowns may pose illiquidity risk to
banks.
The other risk management role of securitization comes in providing banks with
a means to handle their credit risk exposures. When the loans are securitized and
transferred to the SPE, the loans leave the bank balance sheet. This results to the
banks being isolated from the risk of the loans that they have securitized. Many
studies have shown that banks attempt to take advantage of this credit risk
management property of securitization. For example, Minton, Sanders & Strahan
(2004), Bannier & Hänsel (2007) and Panetta & Pozzolo (2010), have all shown that
banks with more risky assets tend to securitize more. However, what may have been
missed in these studies is an examination of what happens as the banks securitize.
That is, these studies have shown that banks securitize, so that they might take some
loans out of their balance sheets and tone down their credit risk exposures. But,
these studies have not shown if the removal of loans through securitization has
1 Loutskina (2011) measures loan portfolio securitizability as the size of the bank’s loan portfolio multiplied by
the depth of the securitization market of the bank’s home economy. 2 Liquid Assets is defined in Loutskina (2011) as marketable securities plus Federal Funds Sold (i.e Reverse
Repurchase Agreements).
3
indeed made banks get rid of risk and has, therefore, reduced the credit risk that
they respectively face. In this paper, we intend to provide the continuation of this
discussion, where we attempt to look at how the loan portfolios of banks change, in
terms of size and composition, with securitization. Moreover, we also investigate
what such change (if any) implies on the risks and returns that the banks eventually
face.
Using data on 129 member banks of the Federal Deposit Insurance Company
(FDIC), we initially look at how securitization may affect the size of the bank’s loan
portfolio. We observe that banks that securitize more (i.e. banks with medium to
high levels of securitization activity) have relatively bigger loan portfolio size, than
those that securitize less. While these findings may be due to the additional funding
provided by securitization3, such explanation may not be enough. Having more
funds (through securitization) does help in increasing a bank’s loan portfolio, but
matters on risk exposures also come at play. Thus, we argue that the increase in the
loan portfolio size that come with securitization could as well be driven by
securitization giving banks a means of getting rid of risk, that makes the banks more
flexible in granting and holding loans. In this sense, we may then consider that the
credit risk management property of securitization can increase the banks’ loan
portfolio size.
Subsequently, we look at the composition of the banks’ respective loan
portfolios, alongside their securitization activities. We view loan portfolio
composition in terms of the share of the different loans classes on the banks’ balance
sheet. We consider five loans classes namely, Real Estate, Commercial & Industrial
(C&I), Consumer, Farm and Others4. A clear observation we make is that the
portfolio share of Consumer Loans is bigger (smaller) for the banks with high (low)
securitization activity, while that of Real Estate Loans is smaller (bigger) among
banks that securitize much (less). This suggests that the portfolio share of Consumer
Loans increases with securitization activity, while that of Real Estate Loans
decreases. In addition, given that Real Estate Loans are the most securitized loans
by banks while Consumer Loans are securitized much less, we also get the
impression that banks may have increased their holdings of Consumer Loans,
through the securitization of their Real Estate Loans.
We attribute the above occurrence to the high risks of Consumer Loans and Real
Estate Loans and to the high positive covariance of risk between the said loans
3 As pointed out by numerous studies such as Cantor & Demsetz (1993), Altunbas, Gambacorta & Marquez-
Ibanez (2007), Goderis, Marsh, Costello & Wagner (2007), Loutskina & Strahan (2008) and Gambacorta &
Marques-Ibanez (2011). 4 The loans class “Others” refers to Acceptances and Receivables discounted by banks.
4
classes. Looking at the default rates of the different loans classes, we observe that
Consumer Loans and Real Estate Loans have the highest default rates, with the
former having higher default rates than the latter. This makes the two said loans
classes the riskiest loans that banks can hold. At the same time, we also observe that
Consumer Loans and Real Estate Loans have the highest positive covariance of
defaults. This means that holding both Consumer Loans and Real Estate Loans
simultaneously may be too risky, which requires the banks to unload one class of
loans when they hold on to the other. Since Consumer Loans are more monitoring
intensive than Real Estate Loans5, banks may have opted to hold on to Consumer
Loans, where holding and closely monitoring this loans class could realize its high
potential interest income (that comes with its observed high risk6). At the same time,
the banks may have also unloaded Real Estate Loans (while holding Consumer
Loans) because Real Estate Loans are easier to securitize, owing to the
collateralization of this loans class7.
With these observations and its accompanying explanations, we may then
consider the idea that banks may have increased their loan portfolios towards more
Consumer Loans, by taking some Real Estate Loans out of their balance sheet
through securitization, which consequently gives them more space to take on the
risky but high-yielding Consumer Loans. This implies that the banks’ employment
of the credit risk management property of securitization may be in the direction of
being exposed to greater risk, while pursuing high returns.
To test our inference, we implement a series of empirical estimations. First, we
look at the relationship between securitization and loan portfolio size, to verify our
initial observation that banks may have engaged in securitization, so that they can
enlarge their loan portfolios. Employing the data on our 129 FDIC member banks,
we find such confirmation through a positive relationship between securitization
5 Consumer Loans are monitoring intensive because these loans are uncollateralized loans to individuals,
where the failure of such a loan gives zero pay-off to the bank (i.e. the bank must monitor the Consumer Loans
to almost surely avoid losses). On the other hand, Real Estate Loans are collateralized loans for residential
home purchases, where the failure of such a loan can still pay-off the bank through its seizure of the collateral
(i.e. the bank does not need to strongly monitor a Real Estate Loan because, should (at worst) the loan fails,
the bank can still receive a pay-off by taking the collateral). 6 As the risk-reward principle posits.
7 The ease of securitizing a loan may depend on the appeal of such loan to serve as the underlying asset which
backs the debt issued in the securitization transaction. This appeal, in turn, rests on the value of the loan itself.
A collateralized loan’s value is derived from the likelihood that it successfully pays-off and from its collateral.
Meanwhile, an uncollateralized loan’s value is derived solely from the said probability of a successful pay-off. A
loan that is collateralized may then be of higher value, as opposed to an uncollateralized one, and is then also
of higher appeal as an underlying asset in a securitization transaction. It thus follows that a collateralized loan,
such as a Real Estate Loan, is easy or easier to securitize (relative to an uncollateralized loan such as a
Consumer Loan).
5
and total loans holdings. Next, we check if this increase in the loan portfolio,
through securitization, is indeed geared towards holding Consumer Loans. Our
results show that securitization is positively related to the portfolio share of
Consumer Loans, establishing the above point. In our following empirical analysis,
we seek to prove that banks have used securitization to be able to take on higher
risk, embodied in the increased holding of Consumer Loans above. Using loan
defaults as indicators of loan portfolio risk, we find that securitization has a positive
effect on the overall loan portfolio risk of our sample banks. This implies that banks
may have exposed themselves to higher risk as they securitize, contrary to risk-
unloading that has been expected of securitization (in principle). Lastly to see if the
banks have done so to get higher returns, we take the banks’ Return on Assets
(ROA) and Return on Equity (ROE) and estimate them against securitization
activity. With securitization being positively related to both the ROA and the ROE,
we come to the conclusion that banks have certainly made use of the credit risk
management property of securitization to engage in greater risk, so that they may
reap high returns.
However, we also find that the above returns achieved through securitization,
though high, are volatile. We draw this after observing a positive effect by
securitization on the volatilities of the ROA and that of the ROE. The implication of
this finding is that although the high risk-taking may be accompanied by high
returns, the instability of such returns may pose a concern to the banks. This concern
could sequentially compromise the banks’ interest in continuing to engage in
securitization. In such a situation, a certain windfall that would convince the banks
to sustain their securitization activities may be necessary. We find this windfall as
we look into the diversification effects of securitization.
With our earlier finding that securitization can lead to the increase in the
portfolio share of Consumer Loans, we consider that with securitization, the
diversification of the banks’ respective loan portfolios may have also changed.
Using a modified Herfindahl-Hirschman Index (HHI) as an indicator for loan
portfolio diversification, we find that securitization activity is positively related to
loan portfolio diversification or that securitization can make the banks’ respective
loan portfolios more diversified. Further analysing this point leads us to find that
the alteration of the loan portfolio composition towards diversification, may also
allow banks to enjoy the beneficial effects of diversification, namely reduced overall
loan portfolio risk and lower returns volatility. Through these positive side-effects,
the concerns on increased risk and unstable returns found earlier could be offset. As
6
such, the banks may be motivated to continue and conveniently take on greater risk
through securitization, and enjoy the accompanying high returns.
Our study makes a number of contributions to the literature on securitization.
First, our work builds up on earlier studies that view securitization as a means for
banks to increase their loan-taking activities. Our study goes further on this point
by showing that securitization may not only increase the size of the bank’s loan
portfolios, but may also change the composition of the loan portfolio towards more
risky loans and diversification. Second and more importantly, our study adds to the
literature on securitization as a risk management tool, where it shows that
securitization may not necessarily be used for absolute risk isolation or reduction.
Instead, our study points out that securitization is a risk management tool in the
sense that it can be used by banks to take on more credit risk in pursuit of high
yields, while also reaping the benefits of loan portfolio diversification. Lastly, our
study also contributes to the ongoing discussions on the value of securitization.
Following its involvement in the recent financial crisis, securitization has had an
unfavourable reputation. By showing that securitization has risk-taking, profit-
augmenting and diversification effects, our study points out that securitization may
still have some significance.
The rest of our paper is organized as follows; Section II provides a short
background on the credit risk management property of securitization and our
preliminary data analysis. Section III provides a discussion of the literature related
to our study. Section IV presents our empirical analyses in two parts. Section V gives
some further analyses involving securitization and loan portfolio diversification.
Section VI concludes.
II. Background and Preliminary Analysis
Among the properties of securitization that motivate banks to engage in such
activity is its provision of a means to manage credit risk. When a bank securitizes, it
transfers the pool of loans to be securitized to an SPE. This movement takes the
loans out of the bank balance sheet, which effectively isolates the bank from the risk
on these loans. Minton, et. al. (2004), Bannier & Hänsel (2007) and Panetta & Pozzolo
(2010) have all demonstrated that the banks that securitize more are those with more
risky assets. These studies imply that banks may have actually taken advantage of
this credit risk management property of securitization.
However, what these past discussions may have left off is an investigation on the
changes that the banks experience, as they securitize. While the studies mentioned
above may have argued the point that banks securitize with the goal of isolating
7
themselves from risky loans, the studies have not, for example, shown if banks do
indeed have less risky loan portfolios as they securitize. We fill in this gap by
looking at the effects of securitization on the banks’ loan portfolio in terms of size
and composition, as well as, on the resulting risk and returns profiles of the banks.
To go about our analyses, we first look at the securitization and loan portfolio
data of the Top 12989 member banks of the FDIC from 2001-201010. We begin by
taking the average securitization activity11 of each our sample banks for our entire
sample period. Using the average securitization activity of each bank, we rank our
banks from the highest securitization activity to the lowest. This ranking allows us
to cut our sample banks into three groups, where banks above the 67th percentile of
the ranking are considered as the banks that engage much in securitization (i.e.
High-Securitizers), those between the 67th and 33rd percentile of the ranking are
taken as the medium securitizing banks (i.e. Mid-Securitizers) and, the banks at the
bottom 33rd percentile of the ranking are treated as the banks that securitize the least
(i.e. Low-Securitizers). With this grouping of our sample banks, we can take the
average attributes of their loan portfolios (by group). This gives us a bird’s eye-view
of the relationship between securitization activity and the banks’ loan portfolio,
which we report in Table 1.
We can find in Table 1, that securitization may lead to a larger loan portfolio.
Looking at the Total Loans to Assets ratio, we find that the High and the Mid-
Securitizers have bigger loan portfolios than the Low-Securitizers. On the one hand,
banks that securitize much may be able to have more loans because securitization
provides them with the funds that they need to do so12. However, we must take into
account that the holding of loans is not just about having the funds to loan out, but
also the tolerance or the space to bear the risk posed by the loans. As such, the larger
loan portfolios among the heavy securitizing banks could also be an effect of the
credit risk management property of securitization. That is, since securitization
provides the banks with a vent for loans that they may eventually deem too risky,
banks that are much engaged to securitization may be more accommodating or
flexible in giving loans. Hence, securitization may increase the loan portfolio size of
the banks.
8 In Terms of Assets.
9 We started with the Top 150 FDIC Member banks but due to mergers and closures within our sample period,
as well as, missing reports, our number of sample banks gets trimmed to 129. 10
We begin at 2001:4 since this is the period when the FDIC member banks have started reporting their
Securitization Activities. 11
We measure the securitization activity of our sample banks through its reported Bank Assets Sold &
Securitized normalized to Total Assets. 12
See Note 3.
8
Considering this increase in the loan portfolio associated with securitization, we
look at next how securitization may also change the composition of the loan
portfolio. We view the composition of the loan portfolio in terms of the different
loans classes that the banks hold. We consider five loans classes namely, Real Estate,
C&I, Consumer, Farm and Others13. To able to see the effect of securitization on the
composition of the loan portfolio, we take the portfolio shares of each loans class
among our sample banks. We report the average of these portfolio shares per bank
group in Table 1. From Table 1, the clear observations we can make are that the
portfolio share of Consumer Loans (i.e. Consumer Loans/Total Loans) increases
monotonically with securitization, while that of Real Estate Loans decreases
monotonically with securitization. Given this we find that securitization seems to be
associated with more holdings of Consumer Loans (relative to other loans classes) as
well as lesser holdings of Real Estate Loans (relative to other loans classes).
With these opposing trends between the holdings of Consumers Loans and that
of Real Estate Loans in relation to securitization, we entertain the possibility that
banks may have increased their Consumer Loans holdings through the
securitization of Real Estate Loans. We get support for this idea in Figure 1, where
we find that Real Estate Loans are the loans that have been primarily securitized by
banks, and that banks have been securitizing Consumer Loans less. Figure 1
presents the quarterly share of securitized loans to total loans outstanding of each
loans class from 2001-2010. We can see in Figure 1 that, throughout the period we
consider, banks have been predominantly securitizing Real Estate Loans, while
Consumer Loans have not been securitized as much by banks. In fact, towards the
time when securitization has been a popular activity among banks (i.e. 2006), we see
in Figure 1 that the securitization of Real Estate Loans have had some increase while
that of Consumer Loans have remained flat. In addition, we also find in Figure 1 that
even when securitization has fallen out of favour at about 2009 (due to its
involvement with the subprime crisis), Real Estate Loans is still the leading loans
class that is being securitized.
We attribute this likely scheme of securitizing Real Estate Loans to hold
Consumer Loans to the high risks of these two loans classes and to the high positive
covariance of risk between these two loans classes. In Figure 2, we plot the default
rates of our different loans classes. We can observe in Figure 2 that Consumer Loans
have the highest default rates among the different loans classes, followed by Real
13
See Note 4.
9
Estate Loans14. This means that the said two loans classes may be the riskiest loans
that the banks can hold. Meanwhile, in Table 2, we report the covariances of the
default rates of our different loans classes. We can see that the pair of loans classes
with the highest positive covariance of risk is that of Consumer Loans and Real
Estate Loans. Combining these two observations imply that simultaneously holding
Consumer Loans and Real Estate Loans may be too risky for banks. That is, since
Consumer Loans and Real Estate Loans are the riskiest loans classes that banks can
hold, and that these two loans classes have a very high positive risk covariance,
holding much of the two of them at one time, may exhaust a bank’s tolerance for risk
(i.e. risk limits). Thus, it may be then be necessary for the bank to unload one loans
class (i.e. the Real Estate Loans), when holding the other loans class (i.e. the
Consumer Loans). Considering that Consumer Loans are loans to individuals that
have no collateral, Consumer Loans require more monitoring, where doing so can
realize its high potential returns (associated with its high risk)15. As such, banks may
have chosen to hold more Consumer Loans, as opposed to Real Estate Loans. In
addition, banks may have also securitized Real Estate Loans (as they hold Consumer
Loans) because the collateralization of Real Estate Loans, makes it easier for this
loans class to be securitized, than the uncollateralized Consumer Loans (and even
the other loans classes). The ease of securitizing a loan may depend on the value of
such loan to serve as the underlying asset that backs the debt in a securitization
transaction. Real Estate Loans, by being collateralized, surpass Consumer Loans and
other loans classes in such qualification, because the collateral of Real Estate Loans
pushes up that value16.
From our above discussion, we may construe that securitization must have been
used by banks, not necessarily to totally isolate themselves from risk (as the theory
may expect). Rather, what we have is that banks may have taken advantage of the
credit risk management property of securitization, by using it as a means to gain the
flexibility to take on higher risk and consequently enjoy the high returns promised
by such risk increase. To put this in terms of our observations above, through the
securitization of Real Estate Loans (which takes these loans out of the balance sheet),
banks have managed to gain some space to increase their loan portfolios towards
Consumer Loans. As we find, Consumer Loans is the riskiest loans class that banks
14
We note that we can see in Figure 2 that the default rates of Real Estate Loans may have caught up with
that of Consumer Loans in 2008:2. However, we must consider that this period may have been triggered by an
exogenous shock, namely the bursting of the US Housing Market Bubble. Following the US Housing Market
Collapse, we can see in Figure 2 that Consumer Loans still have relatively higher default rates than Real Estate
Loans. 15
See Notes 5 & 6. 16
See Note 7.
10
can hold implying that the banks’ taking in of Consumer Loans through
securitization, is an increase in risk taking. Moreover, as we point out, banks have
held on to more Consumer Loans to be able to monitor them intensively and realize
the high potential returns that their high risk promises. Given this, the banks’
expansion of its loan portfolio towards risky Consumer Loans through
securitization, may be in the pursuit of high returns.
To confirm our hypothesis we employ our data on the 129 member banks of the
FDIC in a series of estimations. First, we test if securitization does have a positive
relationship with loan portfolio size, to verify our point that securitization has been
used by banks to gain more flexibility in taking on loans. Next, we look at the
relationship between securitization and the portfolio share of Consumer Loans, to
see if the increase in the loan portfolio of the banks has indeed been towards
Consumer Loans. Following this, we investigate if securitization is positively related
to the overall loan portfolio risk of the bank, which will establish that banks have
changed their loan portfolios through securitization, in the direction of taking on
higher risk. Lastly, we examine if securitization is also positively related to the
returns of the bank, to see if the above move of increasing risk exposures has been
motivated by the interest of getting high returns17. However, before we proceed to
these analyses we briefly discuss the literature related to our study.
III. Related Literature
Our study is primarily related to Cebenoyan & Strahan (2004) that looks into bank
engagement in loan sales, which is an analogue to securitization18. In Cebenoyan &
Strahan (2004), it has been observed that banks that actively sell their loans can
increase their risky loans holdings in the form of Commercial Real Estate loans and
C&I loans. In other words, the study has found that loans sales have an effect on the
bank’s loan portfolio, where such portfolio gets to admit more risky loans.
Following this change in the loan portfolio, the study investigates if the bank’s risk
and returns may have been affected. The study’s results show that the volatility of
17
Which may be achieved with the complimentary increased monitoring of the risky exposures 18
Like securitization, the loans are sold to third parties under loans sales and, hence, leave the bank balance
sheet, isolating the bank from the risk of these loans. The difference (among others) in the case of loan sales is
that, the loans are sold directly to investors, which makes the investors’ risk exposures contingent on the loans
that they have respectively bought. On the other hand, in the case of securitization, the loans are sold to the
SPE, which in turn, issues debt backed the loans it has purchased to the investors. In effect, the risk exposures
of the investors, in securitization, rely on the pool of loans that backs the SPE-issued debt. Notwithstanding
this difference, however, loan sales and securitization are, as mentioned, similar in terms of the risk
management property that they provide to banks.
11
loan defaults are not statistically correlated with loan sales, implying that the
increase in risky loan holdings stemming from loan sales does not necessarily
translate to a higher overall risk profile for the bank. At the same time, the study has
observed that loan sales are also not statistically related to the ROA and ROE of the
banks, but are instead negatively related to the volatility of the ROA and that of the
ROE. These suggest that risky loan holdings brought about by loan sales may not
also compromise the bank’s returns and that it may even make the banks’ returns
more stable. The conclusion that may be drawn from the study is thus, that loan
sales serve as a risk management tool for the bank, wherein it allows the bank to
take on higher risk without exacerbating the bank’s overall risk profile and, at the
same time, gives the bank the benefit of more stable returns.
On studies that focus on securitization, our study is related to Dione &
Harchaoui (2003) and Loutskina (2011), where both studies show that securitization
can lead to banks being able to grant and hold more risky loans. In Dione &
Harchaoui (2003) it has been pointed out that increasing levels of securitization tend
to increase the risk-weighted assets to assets ratio among Canadian banks.
Meanwhile, Loutskina (2011) has shown that banks with more securitizable loan
portfolios can continue to lend to the C&I Sector, and even increase this lending
during periods of funding shocks. The said study points out that C&I loans may be
the riskiest and least liquid types of loans that banks can grant.
In addition, our study is also connected with Pavel & Phillis (1987) and Panetta &
Pozzolo (2010), through our later discussion on the diversification effect of
securitization. In Pavel & Phillis (1987), it has been found that banks participate in
loan sales to have more diversified loan portfolios. The said study has shown that
banks that are less diversified are more likely to sell their loans and that the banks
that have done so attain more diversified loan portfolios (as opposed, to those that
have not engaged in loan sales). On the other hand, Panetta & Pozzolo (2010) has
found that banks that securitize, tend to end up with more diverse loan portfolios.
Specifically, the study has observed, that banks that have securitized, have loan
portfolio compositions that are more equally distributed among mortgages, leases
and other loans.
IV. Empirical Analyses
We divide our empirical analyses into two parts. The first part deals with the banks’
usage of securitization to make changes in their loan portfolios, in terms of size and
composition. For the second part of our analysis, we examine if such changes made
12
by banks through securitization are in the interest of taking in greater risk, to reap
high gains.
To review, we implement our empirical analysis using our data on the 129
member banks of the FDIC from 2001-2010. The balance sheet data of our sample
banks are found in their submitted Call Reports that are accessible at the FDIC
website. The Call Reports are available on a quarterly basis.
For all our estimations, we measure securitization activity by taking the sum of
Bank Assets Sold and Securitized reflected in our banks’ Call Reports. We normalize
this figure to Total Assets. This variable for securitization activity, which we shall
note as Securitization, will be our main variable of interest in our empirical analyses.
A. Securitization and the Bank Loan Portfolio
i. Loan Portfolio Size
We begin our empirical analysis with the test on the banks’ usage of securitization to
gain more flexibility in holding more loans. We do this by looking at the relationship
between securitization and loan portfolio size. Our specification is as follows:
itititoit XtionSecuritizaY εβαα +++= −− 111 (1)
Where: itY = Total Loans Holdings or Loan Portfolio Size of bank i at the end of
quarter t; 1−ittionSecuritiza = securitization activity of bank i at the beginning of
quarter t; and 1−itX = vector of bank-specific control variables of bank i at the
beginning of quarter t.
To measure Loan Portfolio Size, we take the reported Total Loans and normalize
it to Total Assets (Total Loans/Assets). We expect that Securitization should be
positively related to Total Loans/Assets (i.e. 01>α ). As we have argued
securitization creates a vent for the banks to rid themselves of risk. With this facility,
a bank may then be more accommodating in granting and holding loans, which
leads to an increase in its loan portfolio size.
To account for the other factors that may have an influence on the banks’ loans
holdings (besides securitization), we add bank-specific control variables in our
specification. The bank-specific factors we consider are Bank Size (measured
through the Log of Assets), Capitalization (measured through the Capital-to-Assets
Ratio) and the size of the bank’s Traditional Funding Base (measured through Core
Deposits/Assets). We also take into account if the bank operates within only one
state or region (Local Bank Dummy) and if the bank is affiliated to a foreign bank
13
holding company (Foreign Bank Dummy), as these characteristics may also have an
influence on the size of the bank’s loan portfolio19.
The summary statistics of the control variables involved in our estimations are
also reported in Table 1. We can observe from our summary statistics that banks that
securitize more, are bigger and better-capitalized. These bank characteristics may
lend support to the banks, as they hold more loans through securitization. We also
observe that banks that securitize more are less dependent on traditional funding.
As we move from the Low-Securitizers group to the High-Securitizers group, we
observe that Core Deposits/Assets is decreasing. This could be due to the other
property of securitization, which is that of being a funding facility20.
In implementing our specification, we employ bank fixed effects (FE) and panel
instrumental variables (IV) regressions. In our IV regressions, we use as instruments
the previous period’s21 Securitization Activity, Bank Size, Capitalization and
Traditional Funding Base. Moreover, we also use as instruments a measure for the
extent of Credit Enhancements22 that a bank offers on its securitization transactions,
as well as, a measure for riskiness of the loans that the banks have securitized23. The
previous period’s Securitization Activity, Credit Enhancements and riskiness of
securitized loans affects our dependent variable (i.e. loan portfolio size), only
through our main explanatory variable24 (i.e. the beginning of the quarter’s
19
Local banks may have less market coverage than national banks and thus may have smaller loan portfolios.
Foreign banks, on the other hand, due to their reputation and larger market coverage, may have larger loan
portfolios. 20
See Karaoglu (2005), Martin-Oliver & Saurina (2008) and Loutskina (2011) for studies on the funding
property of securitization. 21
Given that we use a one-period lag on our independent variables, these instruments mentioned will then
take a two-period lag. 22
Credit Enhancements are special features that banks may offer on their securitization transactions to make
their deals more attractive. For a comprehensive discussion on Credit Enhancements and its various types, see
Lea (2006). In this study, the Credit Enhancements that we consider are Subordinated Securitization and Lines
of Credit. We measure the extent by which our sample banks offer these features by taking the ratio of the
Amount of Credit Enhancements to the Total Amount of Securitized Loans (i.e. Total Amount of Subordinated
Securitization Retained by the Bank/Total Bank Assets Sold and Securitized, Lines of Credit on Securitized
Loans/Total Bank Assets Sold and Securitized). Data on Credit Enhancements are reported in the FDIC Call
Reports. 23
We measure the riskiness of our sample banks’ securitized loans by taking the default rate of the banks’
securitized loans. We calculate the default rate as the ratio of the Amount of Securitized Loans in Default to
the Total Amount of Securitized Loans (i.e. Total Amount of Securitized Loans in Default/Total Bank Assets
Sold and Securitized). 24
The said instrumental variables have no direct effect on the observed loan portfolio size, but may only be
related to it through Securitization Activity (which, as argued, may have a positive relationship on the loan
portfolio size). A bank that has had a large Securitization Activity for the previous period may be able to
securitize more in the following periods, as its previous securitization activity could allow it to build a certain
expertise and reputation in carrying out the transaction (Albertazzi, Eramo, Gambacorta & Salleo (2011)).
Likewise, a bank that offers much Credit Enhancements on its securitization transactions may also be able to
14
Securitization Activity (1−ittionSecuritiza )). Maintenance of our results in the FE
estimations under our IV estimations, will then rule out the issue of reverse causality
between securitization and loan portfolio size.
In addition to using the two above estimation processes, we also execute our
specification under three different periods. The first period we consider
encompasses our entire sample period from 2001:4-2010:4. For the second period, we
consider only 2001:4-2009:2 (which is the period when securitization has been
popularly practiced by banks), while the third period concerns only 2009:3-2010:4
(which is the period when banks have mainly shied away from securitization25). We
differentiate between these periods to be able to establish if, even at the time when
securitization has already been unpopular due to its entanglement with the
subprime crisis, banks can still use it to be able to hold more loans.
Our estimation results are reported in Panel 1 of Table 3 and in Panel 1 of Table
4. Both the FE and IV estimation results are presented in Tables 3 & 4. In Table 3 the
results concern the entire sample period of 2001:4-2010:4, while in Table 4 the results
are differentiated according to when securitization was in favour (2001:4-2009:2) and
when it has been less so (2009:3-2010:4). We find in both FE and IV estimations in
Panel 1 of Table 3 that Securitization is positively related to Loan Portfolio Size, as
expected. Our results show that this effect of Securitization is, as well, statistically
and economically significant. This finding establishes the point that banks may have
used securitization to take in more loans. As we have pointed out, securitization
gives the bank the option to unload loans, should they deem some loans too much to
bear. Given this outlet, banks may then be more flexible in holding loans, which can
inflate their portfolios. In Panel 1 of Table 4, we see that this point applies in both
periods when securitization was popular and when it was otherwise. We find that in
both differentiated periods, the FE and IV estimations still show a positive
relationship between Securitization and Loan Portfolio Size
On the control variables all our results in Panel 1 of Table 3 and Panel 1 of Table
4, show that Log of Assets is negatively related to Total Loans/Assets. Though this
may be an unexpected result, this can be explained by big banks being more
involved in other activities such as fee-based services (e.g. securities underwriting,
securitize more, because the Credit Enhancements serve as guarantees that the securitization transactions
involve good quality underlying assets and that it will pay-off its investors, even if a default in the underlying
assets happen (Thomas (1999), Gorton & Souleles (2005), Ashcraft & Schuermann (2007)) . Meanwhile, a bank
with risky securitized loans may find its securitization activity limited, as its high incidence of defaulting
securitized loans may signal that the underlying assets on its securitization transactions are of poor quality. 25
This period may be observed in Figure 1 as the horizon where there has been a steep drop in the
securitization of Real Estate Loans (as well as Consumer Loans) and that there has been no recovery since then
on in the securitization of loans (irrespective of class).
15
loan syndication, etc.). Involvement in these other activities tends to compromise the
bank’s interest-income driven activities, specifically, loans. Hence, the negative
relationship observed between loan holdings and bank size. On the other hand, the
Capital-to-Assets Ratio (CAR) has come out positively related to total loans
holdings, also in all estimations reflected in Panel 1 of Table 3 and Panel 1 of Table 4.
This result is straightforward, as more capitalized banks are more capable of bearing
loan exposures.
ii. Loan Portfolio Composition
Having established that banks use securitization to expand their loan portfolios, we
proceed to investigate if such, as we have earlier observed, is geared towards
holding more Consumer Loans. To do this, we re-estimate Equation (1) with the
portfolio share of Consumer Loans (Consumer Loans/Total Loans) as dependent
variable. In our estimations, we implement the same estimation techniques and
treatments in the sample period, as in our previous analysis between securitization
and loan portfolio size. We expect that securitization should be positively related to
Consumer Loans/Total Loans (i.e. 01>α ).
From our set of results in Panel 2 of Table 3, we find a confirmation of our above
expectation. Securitization is positively related to Consumer Loans/Total Loans, with
this effect being statistically and economically significant. At the same time, from
our results in Panel 2 of Table 4, we observe that the positive relationship between
Securitization and Consumer Loans/Total Loans holds in both periods when
securitization was highly practiced and when it was not. Going through our control
variables, the robust and economically significant relationships to Consumer
Loans/Total Loans that we find are those on Capitalization and Core
Deposits/Assets. We have in the results of both Panel 2 of Table 3 and Panel 2 of
Table 4 that Capitalization is positively related to Consumer Loans/Total Loans.
This result reiterates the importance of capital adequacy in bearing loan exposures,
especially risky loans such as Consumer Loans. Meanwhile, Core Deposits/Assets is
negatively related to Consumer Loans/Total Loans in all our estimations. This
implies that banks relying much on traditional funding may not grant and hold
much Consumer Loans. This can be explained by the banks creating a mismatch on
their assets and liabilities to manage cash flows, since both Core Deposits and
Consumer Loans are retail in nature26.
26
Consumer Loans and Core Deposits are retail banking activities in the sense that banks transact such
activities with individual clients. In a period when there is a high demand for cash among consumers, a surge
in applications for Consumer Loans might happen alongside deposit withdrawals, creating a cash flow
16
With our findings above, we have managed to prove the first two points of our
hypothesis. These points are that banks have taken advantage of the credit risk
management property of securitization to be able to increase their loan portfolios,
and that this increase is geared towards Consumer Loans. In the following analyses,
we seek to confirm the remaining part of our hypothesis. This is that as the banks
make the said changes in their loan portfolios using securitization, they expose
themselves to higher risk and that this exposure to greater risk has been done to
enjoy high returns (that an increase in risk-taking with intense monitoring
promises).
B. Securitization, Loan Portfolio Risk and Bank Returns
i. Loan Portfolio Risk
To check if banks may have exposed themselves to more risk through securitization,
we look at the relationship between securitization and the overall risk of the bank’s
loan portfolio. We take four indicators of loan portfolio risk namely; a. the Non-
Performing Loans27 to Total Loans Ratio (NPLs/Total Loans); b. the standard
deviation of NPLs/Total Loans28 (SD NPLs); c. the Loan Charge-Offs29 to Total
Loans Ratio (Charge-Offs/Total Loans); and d. the standard deviation of Charge-
Offs/Total Loans30 (SD Charge-Offs).
Given our measures for loan portfolio risk, we do panel estimations of the
following equation:
ititititoit ZXtionSecuritizaY εφβαα ++++= −−− 1111 (2)
Where: itY = Loan Portfolio Risk Indicator of bank i at the end of quarter t;
1−ittionSecuritiza = securitization activity of bank i at the beginning of quarter t; 1−itX =
vector of bank-specific control variables of bank i at the beginning of quarter t; and
1−itZ = vector of shares of each loans class in the loan portfolio of bank i at the
beginning of quarter t.
For the bank-specific control variables, we include the same control variables for
Bank Size and Capitalization plus a control variable for Loan Portfolio Size,
measured by Total Loans normalized to Total Assets (Total Loans/Assets). We
problem. Thus banks, may want to avoid engaging in both retail lending and borrowing activities
simultaneously, or in other words, create a mismatch in their assets and liabilities. 27
Non-Performing Loans are defined as loans that have been past due for 90s days plus loans that have been
in non-accrual status. 28
In calculating SD NPLs, we use NPLs of the four quarters of each year. As a result we experience a reduction
in the number of observations when using this as indicator for loan portfolio risk. The same case applies to SD
Charge-Offs. 29
Loan charge-offs are defined as loans that have been delinquent for at least 120 days. 30
See Note 28.
17
control for Loan Portfolio Size as this variable may have an influence on the
incidence of loan defaults (which we use as indicators for loan portfolio risk). Like in
our previous estimations we also take into account the geographical confinement of
the bank’s operations (Local Bank Dummy) and its affiliation to a foreign bank
holding company (Foreign Bank Dummy).
At the same time, we also include variables controlling for the shares of each
loans class in the banks’ respective loan portfolios (vector-Z). We include this set of
control variables, considering that each loan class have varying risk and may
therefore have different impacts on the incidence of loan defaults. To avoid perfect
collinearity, we exclude the portfolio share of Other Loans (i.e. Other Loans/Total
Loans).
Going back to Table 1, we see in our summary statistics that our indicators of
loan portfolio risk are generally higher for banks that securitize more (i.e. those in
the Mid and High-Securitizers Group), than those that securitize less (i.e. those in
the Low-Securitizers Group). This gives us the impression that overall loan
portfolio risk may increase with securitization. In Table 5, we get a verification of
this notion where we find a positive relationship between Securitization and all our
measures of loan portfolio risk. As reported in Table 5, securitization is statistically
and economically significant in most of our estimations except that of SD Charge-
Offs. These results affirm our point that banks have used securitization not
necessarily to totally isolate themselves from risk, but rather to be able to bear
greater risk.
On the control variables, Banks Size, Capitalization and Loan Portfolio Size are
chiefly positively related to loan portfolio risk. Bigger and more capitalized banks as
well as banks with larger portfolios do tend to take on more risk. Hence, such
variables are bound to be positively related with loan portfolio risk. Meanwhile, the
respective portfolio shares of the different loans classes are also mostly positively
related to all the measures of loan portfolio risk. This result is immediate, as every
loans class must pose some risk to the loan portfolio of the bank.
ii. Bank Returns
Finding that banks may have securitized towards increasing their loan portfolio risk,
we go to our other point, where this move may have been stirred by the banks
wanting to realize the high potential returns of such risky loan exposures (with more
intense monitoring). Should this be the case, then securitization activity must be
associated with high returns for the banks. We measure bank returns in two
dimensions namely the actual returns of the bank and the stability of these returns.
18
Our measures for the actual bank returns are the ROA and the ROE, while for the
stability of bank returns, our indicators are the respective standard deviations of the
ROA and ROE (SD ROA and SD ROE)31. The higher the SD ROA or the SD ROE, the
less stable are the bank returns.
Looking at the summary statistics of our variables for bank returns in Table 1, we
find some evidence that the increase in risk taking through securitization may be
accompanied by high bank returns. We observe that the ROA and the ROE for High-
Securitizers are markedly higher than that of Mid-Securitizers and Low-Securitizers.
However, it also seems that these high returns may not necessarily be
complemented with more stable bank returns, as the SD ROA and the SD ROE
appear higher for the groups of banks that securitize more.
To verify our observations, we estimate our variables for bank returns against
the same independent variables we used in the estimations for loan portfolio risk
(i.e. Equation 2). The results are presented in Table 6. In Panels 1 & 2 of Table 6, we
find that Securitization is positively related to both the ROA and ROE and is also
statistically and economically significant in both estimations. These findings along
with the earlier ones on loan portfolio risk, confirm our point that banks may have
used securitization to take on higher risk and enjoy high returns. This is emphasized
with the portfolio share of risky Consumer Loans being also positively related to the
ROA, while the portfolio shares of the other loans classes have coefficients with
negative signs. On the other control variables, we have that bank size and
capitalization are negatively related to bank returns, pointing out that big and more
capitalized banks are not necessarily the most profitable. This may be once again
due to the tendency of most of these banks to go into other activities where these
activities may not be very high-yielding (e.g. fee-based services). Loan portfolio
size, on the other hand, is positively related to both ROA and ROE. These results are
driven by the large interest income that is derived from a big loan portfolio.
Meanwhile, Panels 3 & 4 show that securitization is positively related to the SD
ROA and the SD ROE, or that securitization can lead to unstable bank returns. These
results may be due to the risky source of the high returns provided by securitization,
which is that of increasing the portfolio share of risky loans. The implication of these
findings is that, although the banks’ increase in risk exposures through
securitization may yield high gains, these gains may be unstable (owing to the high
31
Like in the case of SD NPLs and SD Charge-Offs, the SD ROA and the SD ROE are calculated using the ROA
and the ROE of the four quarters of each year. As a result we experience a reduction in the number of
observations for these particular estimations.
19
risk exposures itself). Given such, the reward of high returns for higher risk may not
be enough to convince the banks to sustain their engagement in securitization. Thus,
some form of windfall might be necessary to persuade banks to continue to
securitize and carry on with the strategy above. We find such windfall in the
complementary diversification effect of securitization, which we discuss in the
following section.
V. Further Analyses
A. Securitization and Loan Portfolio Diversification
Our earlier finding which shows that securitization shuffles the composition of the
bank’s loan portfolio towards more Consumer Loans, brings the possibility that
securitization may also have an impact on the diversification of the bank’s loan
portfolio. Given that diversification has implications on the risk of the bank’s loan
portfolio, as well as its returns, we look into this issue. The classical hypothesis of
diversification posits that by spreading a portfolio into different assets (or, in our
case, loans) with uncorrelated risks, the risks may cancel each other out. As such,
diversification may reduce the overall risk of the loan portfolio and that the risk-
reduction could also lead to more stable returns. In Acharya, Hasan & Saunders
(2002), this hypothesis has been partly verified by results that among Italian banks,
the banks that diversify their loans on different sectors have lower NPLs.
If we look back at our summary statistics in Table 1, we may observe that
securitization, to some degree, could lead to a more diversified loan portfolio. As we
move from the Low-Securitizers Group to the High-Securitizers Group, we see a
relatively more even distribution in the portfolio shares of each loans class. For
example, Low-Securitizers, on average, hold a large chunk of their loans in Real
Estate Loans at 62%, followed by C&I Loans at 24% and only a small portion of
Consumer Loans at 8%. In contrast, High-Securitizers, have a relatively more
balanced loan portfolio, where on average, the portfolio share of Real Estate Loans is
at 45%, while that of the C&I Loans is at 17% and that of Consumer Loans is at 26%.
To further investigate this point, we take an indicator for loan portfolio
diversification given by the inverse of the Herfindahl-Hirschman Index ( HHI−1 )32.
The intuition of this index is that the higher it is, the more diverse is the loan
portfolio. The average HHI−1 of our different bank groups are reported in Table 1.
Both Mid-Securitizers and High-Securitizers have fairly higher average
diversification indicators than the Low-Securitizers. Given such, we move to
32
Please see the Appendix for a discussion on the computation of the Herfindahl-Hirschman Index.
20
confirm this observed case of securitization leading to loan portfolio diversification.
To do so, we implement a panel estimation on itHHI−1 against securitization. In this
estimation, we include the same control variables used in Equation (1). At the same
time, we add a one-period lagged value of the same indicator for loan portfolio
diversification (1
1 −− itHHI ). The purpose of adding 1
1 −− itHHI is to control for the
likelihood that the bank may have already been diversifying previously (perhaps as
a way to manage risk).
Our estimation results reported in the first row of Table 733 shows that
securitization may actually lead to loan portfolio diversification, with Securitization
being positively related to HHI−1 . We note that Securitization is statistically and
economically significant in relation to the loan portfolio diversification indicator,
even while considering for the strong reinforcing effect of the previous extent of loan
portfolio diversification (1
1 −− tHHI ).
B. Securitization, Loan Portfolio Diversification, Risk and Returns
Given the result above, we next examine if the bank reaps the known benefits of
diversification, namely lower overall loan portfolio risk and more stable returns. In
doing this, we estimate our measures of loan portfolio risk and our measures for
bank returns volatility against our diversification indicator (1
1 −− itHHI )34, while
controlling for securitization activity. We also include the control variables we have
used in the estimation of Equation (2), except for the portfolio shares of the different
loans classes, as these shares may already be accounted for by the diversification
indicator35.
Looking at our results on the second to the seventh rows of Table 736, we find
that the banks do enjoy the beneficial effects of diversification. In Panel 1, we can see
that 1
1 −− tHHI is negatively related to our measures of loan portfolio risk, indicating
that diversification lowers overall loan portfolio risk. At the same time, we can
observe that the diversification indicator is negatively related to the SD ROA and the
SD ROE, which means that diversification may stabilize bank returns. Meanwhile, in
Panel 2 we see the same relationship we have found earlier between securitization
and our variables for loan portfolio risk and bank returns stability. In the case of
33
For purposes of brevity, we report in Table 7 only the coefficients of our variables of interest namely
Securitization and the one-period lag of the diversification index (1-HHIt-1). 34
The loan portfolio diversification indicator for this set of estimations takes a one-period lag, since we use
beginning of quarter values for our independent variables, as in the earlier estimations. 35
This is because the calculation of the diversification index is based on the portfolio share of each loans class
in the bank’s loan portfolio. 36
See Note 33.
21
NPLs and SD NPLs, we see that the negative impact of1
1 −− tHHI is more
economically significant than the positive impact of Securitization. These suggest that
the diversification benefit of reduced overall loan portfolio risk, may effectively
temper the increased risk brought about by securitization that results from the
greater portfolio share of risky loans. Likewise, the negative effect of 1
1 −− tHHI on SD
ROE is more economically significant than the positive effect of Securitization. This
points out that the instability of returns from securitization may also be mitigated by
the diversification benefit of stable returns. Given these, we can consider that the
diversification benefits could offset the undesirable effects of securitization. As such,
banks receive the needed windfall to continue to engage in securitization with the
interest of taking on higher risk, in pursuit of high returns.
VI. Concluding Remarks
Securitization provides banks with a means to isolate themselves from risky loans. A
number of studies have shown that banks may take advantage of this credit risk
management property of securitization, where banks with riskier assets tend to
securitize more. We expound on this point by looking at what happens to the banks’
risk exposures as they securitize.
We have observed that when securitizing, banks have larger loan portfolios and
that this increase in loan portfolio size is geared towards having more risky assets.
At the same time, we have found that these changes in the banks’ loan portfolios,
through securitization, entail an increase in the overall risk of the bank’s loan
portfolios as well as high returns. Given these, we construe that as banks securitize
and use the credit risk management property of securitization, what occurs is not
necessarily a total isolation from risk (as what theory may predict). Instead, what
happens is a taking on of greater risk among the securitizing banks, to achieve high
returns.
A concern on the high returns from securitization, however, is that we have also
found these to be unstable. This may due to the fact that the said returns are derived
from the high exposure to risky loans. Nevertheless, this concern, as well as, the
implications of having more risky assets is tempered through the benefits of having
a more diversified loan portfolio, which comes as a side-effect of the engagement in
securitization. As banks use securitization to increase their exposure to risky assets,
they also end up diversifying their loan portfolios that provide them the
diversification benefits of reduced overall loan portfolio risk and more stable
returns. With these diversification benefits offsetting the above issues, banks find the
22
motivation to keep on securitizing to increase their risk exposures, and enjoy the
accompanying high rewards for taking on risky assets.
From our findings, we conclude that the credit risk management property of
securitization goes beyond the mere offloading of credit risk, by taking loans out of
the bank balance sheet. Rather, the credit risk management role of securitization
goes all the way to the structure of the bank’s loan portfolio or the bank’s choice of
assets, where this choice of assets is geared towards more risk-taking, to achieve
better returns.
Reference List
Acharya, V.V., I. Hasan & Saunders, A. 2002. Should banks be diversified? Evidence from individual bank loan portfolios. Bank for International Settlements (BIS) Working Papers, No. 118.
Albertazzi, U., G. Eramo, L. Gambacorta, & Salleo, C. 2011. Securitization is Not that Evil After All. Bank for International Settlements (BIS) Working Papers. No. 341. Altunbas, Y., L. Gambacorta & Marques-Ibanez, D. 2007. Securitization and the Bank Lending Channel. European Central Bank (ECB) Working Papers, No. 838.
Ashcraft, A.B. & T. Schuermann. 2007. Understanding the Securitization of Subprime Mortgage Credit. Mimeo. Bannier, C.E. & D.N. Hänsel. 2007. Determinants of Banks’ Engagement in Loan Securitization. Conference on the Interaction of Market and Credit Risk. Cabiles, N.A.S. 2011. Hedging Illiquidity Risk through Securitization: Evidence from Loan
Commitments. Mimeo. Cantor, R. & R. Demsetz. 1993. Securitization, Loan Sales and the Credit Slowdown. Federal Reserve Bank of New York Quarterly Review, Summer 1993, 27-37.
Cebenoyan, A.S. & P. Strahan. 2004. Risk Management, Capital Structure and Lending at Banks. Journal of Banking and Finance, 28, 19-43.
Dione, G. & T.M. Harchaoui. 2003. Banks’ Capital, Securitization and Credit Risk: An Empirical Evidence for Canada. HEC Montreal Working Paper No. 03-01.
Gambacorta, L. & Marques-Ibanez, D. 2011. The bank lending channel: Lessons from the crisis. BIS Working Papers, No. 345. Goderis, B., I.W. Marsh, J.V. Castello & Wagner, W. 2007. Bank behaviour with access to credit risk
transfer markets. Bank of Finland Research Discussion Papers 4. Gorton, G. & N.S. Souleles. 2005. Special Purpose Vehicles and Securitization. Mimeo.
Kamp, A., A. Pfingsten, C. Memmel & Bher, A. 2007. Diversification and the Banks’ Risk-Return Characteristics-Evidence from Loan Portfolios of German Banks. Deutsche Bundesbank Discussion Paper Series 2, 5.
23
Kamp, A., A. Pfingsten & D. Porath. 2006. Do banks diversify loan portfolios? A tentative answer based on individual bank portfolios. Deutsche Bundesbank Discussion Paper Series 2, 3.
Karaoglu, N.E. 2005. Regulatory Capital and Earnings Management in Banks: The Case of Loan Sales and Securitizations. FDIC Center for Financial Research Working Paper No. 2005-05.
Lea, M. 2006. Securitization: A Primer on Structures and Credit Enhancement. Mimeo. Loutskina, E. 2011. The Role of Securitization in Bank Liquidity Funding and Management. Journal of Financial Economics, 100, 663-684.
Loutskina, E. & P. Strahan. 2008. Securitization and the Declining Impact of Bank Financial Condition on Loan Supply: Evidence from Mortgage Acceptances. Financial Cycles, Liquidity and Securitization Conference Proceedings, International Monetary Fund (IMF).
Martin-Oliver, A. & J. Saurina. 2008. Why do banks securitize assets? International Forum on
Research in Finance Proceedings. Minton, B., A.B. Sanders & Strahan, P.E. Securitization by Banks and Finance Companies: Efficient Financial Contracting or Regulatory Arbitrage? Mimeo.
Panetta, F. & A.F. Pozzolo. 2010. Why do banks transfer credit risk? Bank-level evidence from over one hundred countries. Mimeo. Pavel, C. & Phillis, D. 1987. Why commercial banks sell loans: An empirical analysis. Federal Reserve Bank of Chicago, Economic Perspectives. Stiroh, J. 2004. Do Community Banks Benefit from Diversification? Journal of Financial Services Research, 25(2), 135-160.
Thomas, H. 1999. A Preliminary look at gains from asset securitization. Journal of International Financial Markets, Institutions and Money 9 (1999), 321-333.
24
Table 1. Summary Statistics
Volume of Bank Assets Sold & Securitized/Assets
(Securitization)
Sampling Period: 2001:4-2010:4,
No. Of Banks: 129, No. of Banks in
Each Group: 43
Low-Securitizers
(Sec ≤ 33rd
Percentile)
Mid-Securitizers
(33rd
Percentile<Sec≤
67th Percentile)
High-Securitizers
(67th Percentile<
Sec)
1 2 3
Total Loans/Assets 59.929 64.019 64.126
Real Estate Loans/Total Loans 62.434 57.585 45.991
C&I Loans/Total Loans 23.668 20.940 17.047
Consumer Loans/Total Loans 8.230 9.637 26.774
Farm Loans/Total Loans 0.554 0.425 0.588
Other Loans/Total Loans 5.185 11.441 9.633
Log of Assets 16.338 16.674 17.522
CAR 10.266 11.401 12.374
Core Deposits/Assets 58.062 51.734 43.220
1-HHI 0.433 0.454 0.453
NPLs/Total Loans 1.310 2.106 2.025
SD NPLs 0.290 0.523 0.369
Charge-Offs/Total Loans 0.563 0.920 1.667
SD Charge-Offs 0.168 0.218 0.205
ROA 0.880 0.769 1.180
SD ROA 0.200 0.425 0.394
ROE 9.809 7.613 10.576
SD ROE 2.039 3.830 3.692
Balance sheet data are taken from the FDIC Quarterly Call Reports. 1-HHI is the inverse of the Herfindahl-
Hirschman Index, serving as indicator for loan portfolio diversification. The standard deviations (SDs) are
calculated for each year using quarterly values.
25
Figure 1. Share of Securitized Loans to Total Loans Outstanding by Loans Class (2001-2004)
The share of Securitized Loans to Total Loans Outstanding of each Loans Class is calculated as the ratio of the
Amount of Securitized Loans of each Loans Class to the Total Amount of Loans of each Loans Class (e.g. Share
of Securitized Real Estate Loans = Amount of Real Estate Loans Securitized/Total Real Estate Loans
Outstanding). Data used for the calculation of figures plotted above are taken from the Flow of Funds Account
of the United States, Federal Reserve Board.
26
Figure 2. Default Rates by Loans Class (2001-2004)
The default rate of each Loans Class is calculated as the ratio of the Amount of Loans of each Loans Class that
is in Default to the Total Amount of Loans of Each Loan Class (e.g. Default Rate of the Real Estate Loans=Real
Estate Loans in Default/Total Amount of Real Estate Loans). The default rates are calculated based on the
aggregate values for the entire Commercial and Thrift Banking System under the FDIC. Data on aggregate
values for the Commercial and Thrift Banking System are accessible through the Statistics on Depository
Institutions available at the FDIC website.
Table 2. Covariance Matrix of Defaults among Different Loans Classes (2001-2004)
Farm C&I Consumer Others
Real Estate 0.1436186 0.0604956 0.1908590 0.0314621
Farm
0.0389255 0.0779504 0.0036428
C&I
0.0361855 0.0029371
Consumer 0.0127283
The covariances are calculated using the default rates in Figure 2. The range for
the calculation of the covariances is based in the sample period of 2001:4-
2010:4.
27
Table 3. Securitization and the Bank Loan Portfolio (Entire Period)
Dependent Variable
Total Loans/Assetst Consumer Loans/Loanst
1 2
Sampling Period: 2001:4-
2010:4 FE IV FE IV
Independent Variables
Securitizationt-1 0.081*** 0.086*** 0.428*** 0.475***
(8.706) (6.950) (32.356) (36.663)
Log of Assetst-1 -0.047*** -0.038*** -0.001 0.001
(-20.243) (-6.696) (-0.523) (0.015)
CARt-1 0.476*** 0.628*** 0.354*** 0.805***
(13.961) (11.481) (12.870) (14.003)
Core Deposits/Assetst-1 -0.001 0.019 -0.058*** -0.091***
(-0.071) (1.259) (-11.312) (-5.742)
Local Bank Dummy Yes Yes Yes Yes
Foreign Bank Dummy Yes Yes Yes Yes
No. of Observations 3982 3982 3982 3982
No. of Banks 129 129 129 129
R-Squared 0.378 0.167 0.435 0.426
The main independent variable for these estimations is Securitization measured as the Total Bank Assets Sold
and Securitized normalized to Total Assets. The dependent variable in Panel 1 is Loan Portfolio Size measured
as the ratio of Total Loans to Total Assets. The dependent variable in Panel 2 is the ratio of Consumer Loans to
Total Loans. FE refers to the Bank Fixed Effects estimation results, while IV refers to the Panel Instrumental
Variables estimation results. In the IV estimations, the instruments used are the previous period’s (i.e. two-
period lag) Securitization, Log of Assets, CAR, Core Deposits/Assets, Total Amount of Subordinated
Securitization Retained by the Bank/Total Bank Assets Sold and Securitized, Lines of Credit on Securitized
Loans/Total Bank Assets Sold and Securitized, and Total Amount of Securitized Loans in Default/Total Bank
Assets Sold and Securitized. Items in parenthesis report the t-statistics. * denotes significance at the 10% level,
** at the 5% level and, ***at the 1% level. All regressions include an intercept.
Tab
le 4. Sec
uritiza
tion and the Ban
k L
oan
Portfo
lio (D
iffe
rentiat
ed P
erio
ds)
De
pe
nd
en
t V
ari
ab
le
To
tal
Loa
ns/
Ass
ets
t C
on
sum
er
Loa
ns/
Loa
ns t
1
2
20
01
:4-2
00
9:2
2
00
9:3
-20
10
:4
20
01
:4-2
00
9:2
2
00
9:3
-20
10
:4
F
E
IV
FE
IV
F
E
IV
FE
IV
Ind
ep
en
de
nt
Va
ria
ble
s
S
ecu
riti
zati
on
t-1
0.0
71
**
*
0.0
69
**
*
0.1
39
**
*
0.2
88
**
*
0.4
89
**
*
0.4
89
**
*
0.2
82
**
*
0.4
67
**
*
(7.6
35
) (5
.39
9)
(13
.52
6)
(4.0
75
) (4
8.1
21
) (3
8.1
18
) (5
.17
6)
(5.2
94
)
Log
of
Ass
ets
t-1
-0.0
43
**
*
-0.0
31
**
*
-0.0
58
**
*
-0.0
58
**
*
-0.0
07
**
*
-0.0
04
0
.01
9*
**
0
.01
2
(-1
6.9
12
) (-
5.0
01
) (-
18
.60
6)
(-3
.35
7)
(-3
.49
8)
(0.6
42
) (5
.19
4)
(0.5
83
)
CA
Rt-
1
0.4
97
**
*
0.6
60
**
*
0.5
18
**
*
0.5
31
**
0
.38
4*
**
0
.71
5*
**
0
.59
8*
**
1
.05
7*
**
(1
4.6
62
) (1
1.1
28
) (7
.06
6)
(2.6
14
) (1
2.8
54
) (1
2.1
00
) (9
.94
9)
(4.1
74
)
Co
re D
ep
osi
ts/A
sse
tst-
1
-0.0
02
0
.01
0
0.0
14
0
.02
4
-0.0
62
**
*
-0.0
73
**
*
-0.2
42
**
*
-0.2
62
**
*
(-
0.1
85
) (0
.62
0)
(0.8
91
) (0
.52
9)
(-1
2.0
75
) (-
4.3
83
) (-
17
.94
6)
(-4
.70
0)
Lo
cal
Ba
nk
Du
mm
y Y
es
Ye
s Y
es
Ye
s Y
es
Ye
s Y
es
Ye
s
Fo
reig
n B
an
k D
um
my
Ye
s Y
es
Ye
s Y
es
Ye
s Y
es
Ye
s Y
es
No
. o
f O
bse
rva
tio
ns
32
75
3
27
5
70
7
70
7
32
75
3
27
5
70
7
70
7
No
. o
f B
an
ks
1
29
1
29
1
29
1
29
1
29
1
29
1
29
1
29
R-S
qu
are
d
0.5
58
0
.18
1
0.7
24
0
.10
0
0.5
58
0
.47
8
0.8
65
0
.24
5
29
Table 4 (Previous Page) Notes: The main independent variable is Securitization measured as the Total Bank
Assets Sold and Securitized normalized to Total Assets. The dependent variable in Panel 1 is Loan Portfolio Size
measured as the ratio of Total Loans to Total Assets. The dependent variable in Panel 2 is the ratio of
Consumer Loans to Total Loans. Estimations for the dependent variables are differentiated into two periods,
namely 2001:4-2009:2 and 2009:3-2010:4. The period of 2001:4-2009:2 pertain to the period when
securitization has been highly practiced by banks. On the other hand, the period 2009:3-2010:4 refers to the
period when banks have engaged less in securitization, as the activity has fallen out of favour due to its
involvement with the subprime crisis. FE refers to the Bank Fixed Effects estimation results, while IV refers to
the Panel Instrumental Variables estimation results. In the IV estimations, the instruments used are the
previous period’s (i.e. two-period lag) Securitization, Log of Assets, CAR, Core Deposits/Assets, Total Amount
of Subordinated Securitization Retained by the Bank/Total Bank Assets Sold and Securitized, Lines of Credit on
Securitized Loans/Total Bank Assets Sold and Securitized, and Total Amount of Securitized Loans in
Default/Total Bank Assets Sold and Securitized. Items in parenthesis report the t-statistics. * denotes
significance at the 10% level, ** at the 5% level and, ***at the 1% level. All regressions include an intercept.
30
Table 5. Securitization and Loan Portfolio Risk
Dependent Variable
NPLst Charge-Offst SD NPLst SD Charge-
Offst
Sampling Period: 2001:4-2010:4 1 2 3 4
Independent Variables
Securitizationt-1 0.786*** 0.917*** 0.141*** 0.023
(10.624) (8.240) (4.003) (0.887)
Log of Assetst-1 0.917*** 0.326*** 0.100*** 0.021***
(22.872) (17.296) (5.400) (2.709)
CARt-1 1.670*** 0.366 0.723*** 0.092
(5.409) (1.503) (4.746) (0.907)
Loans/Assetst-1 1.045*** 0.882*** 0.108* 0.046
(9.145) (12.561) (1.698) (1.503)
Real Estate Loans/Loanst-1 1.276*** 0.224*** 0.145* 0.047
(9.181) (3.025) (1.913) (1.293)
C&I Loans/Loanst-1 1.099*** 0.745*** 0.156* 0.154***
(7.304) (7.165) (1.681) (3.428)
Consumer Loans/Loanst-1 0.369** 2.579*** -0.135* 0.218***
(2.445) (20.083) (-1.648) (4.964)
Farm Loans/Loanst-1 4.463*** -0.236 0.841 0.296
(4.678) (-0.279) (1.609) (0.819)
Local Bank Dummy Yes Yes Yes Yes
Foreign Bank Dummy Yes Yes Yes Yes
No. of Observations 3982 3982 971 971
No. of Banks 129 129 129 129
R-Squared 0.230 0.317 0.092 0.071
The dependent variable in Panel 1 is Non-Performing Loans normalized to Total Loans. Non-Performing
Loans are defined as loans that have been past due for 90 days plus loans that have been in non-accrual
status. In Panel 2, the dependent variable is Charge-Offs normalized to Total Loans. Loans are Charged-Off
if they are delinquent for the past 120 days. The dependent variable in Panel 3 is the standard deviation
(SD) of Non-Performing Loans/Total Loans, while in Panel 4, the dependent variable is the SD of Charge-
Offs/Total Loans. The respective SDs are calculated for each year using quarterly values. Items in
parenthesis report the t-statistics. * denotes significance at the 10% level, ** at the 5% level and, ***at the
1% level. All regressions include an intercept.
31
Table 6. Securitization and Bank Returns
Dependent Variable
ROAt ROEt SD ROAt SD ROEt
Sampling Period: 2001:4-2010:4 1 2 3 4
Independent Variables
Securitizationt-1 0.858*** 4.935*** 0.251*** 1.672***
(9.264) (6.208) (4.640) (4.371)
Log of Assetst-1 -0.285*** -2.875*** -0.008 -0.031
(-16.017) (-15.165) (-0.737) (-0.265)
CARt-1 -1.177*** -67.438*** 0.098 -9.307***
(-4.581) (-30.092) (0.572) (-7.489)
Loans/Assetst-1 0.346*** 5.053*** 0.236*** 0.649
(4.514) (6.207) (4.843) (1.413)
Real Estate Loans/Loanst-1 -0.817*** -11.627*** -0.357*** -2.589***
(-8.100) (-11.518) (-4.445) (-3.817)
C&I Loans/Loanst-1 -0.148 -7.164*** -0.239** -3.570***
(-1.123) (-5.213) (-2.340) (-4.014)
Consumer Loans/Loanst-1 0.371*** -0.901 -0.222** -1.381**
(3.102) (-0.818) (-2.461) (-1.992)
Farm Loans/Loanst-1 -2.211*** -18.823*** -0.792 -1.222
(-3.316) (-2.877) (-1.533) (-0.444)
Local Bank Dummy Yes Yes Yes Yes
Foreign Bank Dummy Yes Yes Yes Yes
No. of Observations 3982 3982 971 971
No. of Banks 129 129 129 129
R-Squared 0.178 0.301 0.098 0.096
The dependent variable in Panel 1 is the ROA, while in Panel 2, the dependent variable is the ROE. The
standard deviations of the ROA and the ROE stand as dependent variables and measures of bank returns
stability in Panels 3 and 4, respectively. The respective SDs are calculated for each year using quarterly
values. Items in parenthesis report the t-statistics. * denotes significance at the 10% level, ** at the 5% level
and, ***at the 1% level. All regressions include an intercept.
32
Table 7. Securitization, Loan Portfolio Diversification, Risk and Returns
Independent Variables
No. of
Obs.
No. of
Banks
R-
Squared
1-HHIt-1 Securitizationt-1
Sampling Period: 2001:4-
2010:4 1 2
Dependent Variables
1-HHIt 0.855*** 0.019*** 3982 129 0.976
(107.307) (3.832)
NPLst -0.782*** 0.286*** 3982 129 0.263
(-4.098) (4.576)
Charge-Offst -0.125** 1.730*** 3982 129 0.219
(-2.465) (13.965)
SD NPLst -0.094* -0.017 971 129 0.042
(-2.428) (-0.502)
SD Charge-Offst -0.051* 0.080*** 971 129 0.043
(-1.787) (2.978)
SD ROAt -0.160*** 0.278*** 971 129 0.111
(-4.786) (5.558)
SD ROEt -2.129*** 1.981*** 971 129 0.137
(-5.254) (5.401)
The dependent variables are on the first column (leftmost panel) of the Table. 1-HHI is the indicator of loan
portfolio diversification, measured as the inverse of the Herfindahl-Hirshcman Index (HHI). For brevity, only
the main independent variables, 1-HHIt-1 (in Panel 1) and Securitizationt-1 (in Panel 2) are reported. Items in
parenthesis report the t-statistics. * denotes significance at the 10% level, ** at the 5% level and, ***at the 1%
level. All regressions include an intercept.
33
Appendix: The Herfindahl-Hirschman Index (HHI)
The HHI is originally a measure of market dispersion (or concentration), calculated
using the market shares of a set competing firms. However, the concept of the HHI
has been also applied to create an indicator for loan portfolio diversification. Studies
that have used the HHI as an indicator for loan portfolio diversification include
Acharya, et. al. (2002), Stiroh (2004), Kamp, Pfingsten & Porath (2006) and Kamp,
Pfingsten, Memmel & Behr (2007). The HHI (as a loan portfolio diversification
measure) is calculated as:
∑=
=n
i
iB qHHI1
2 (1A)
Where: BHHI is the HHI or loan portfolio diversification indicator of bank B,
idenotes a particular loan segment in bank B’s portfolio, n is the number of
segments in bank B’s loan portfolio, and
∑=
=n
i
i
i
i
Q
1
or the share of a certain loan
segment i, in bank B’s loan portfolio.
In our study, we have five loan segments, which are the five loans classes we
consider, namely Real Estate Loans, C&I Loans, Consumer Loans, Farm Loans and
Other Loans. The HHI takes a value between n/1 and 1, where a value equal to 1
means that the bank’s loans are of only one segment or that there is full
concentration (i.e. no diversification). On the other hand, an HHI equal to n/1 ,
implies equal shares among the different loan segments in bank B’s portfolio or full
diversification. For our analysis, it is more convenient to take and use the inverse of
the HHI that is, HHI−1 . In this way, we have that a higher (lower) HHI−1 indicates
a more diverse (concentrated) loan portfolio for the bank.